IEEE Consumer Electronics Society Game On: Atari Video Computer System – Bring the story home Joe Decuir, IEEE Fellow IEEE CES Board of Governors 2015-2017 CES Distinguished Lecturer 2013-2016 IEEE Consumer Electronics Society Agenda Bring the story home Early video games The market problem to solve in 1976 How the 6502 changed things Turning a prototype into a product Game design: Tank Brilliant game designers What we learned Resources 2 IEEE Consumer Electronics Society Entertainment: in the network, and then at home IEEE Consumer Electronics Society Entertainment in 1970 When I started the college, the stories were all controlled by the public media companies: Movie companies made films and showed them in theaters TV networks provided news and assorted shows, punctuated by commercials, which we could watch at home Musicians recorded songs, which we could listen to on the radio or on vinyl records at home – I was lucky enough to see some of them in concert A few of us also made it to live theaters (with actors) We – the customers - did not control the story 4 IEEE Consumer Electronics Society First Video Games Ralph Baer was a pioneer, recognizing that it was possible to bring entertainment home. He imagined a machine which allowed electronic gaming on a “Brown Box” in a family home. Ralph was unlucky – he worked in defense Ralph Baer His employer licensed his in his basement design to Magnavox – as lab, 2014 the Odyssey 5 IEEE Consumer Electronics Society Video Games in arcades Video games were known from experiments on university computers Entrepreneurs like Nolan Bushnell experimented with bringing the gaming experience to arcades, starting with Computer Space Nolan was lucky to witness a demonstration of the original Magnavox Odyssey (Brown Box) Al Alcorn re-implemented the game idea as an arcade game, named Pong – This was wildly successful In 1975, they brought out a home version – another success 6 IEEE Consumer Electronics Society Microprocessor Revolution In the mid-20th century, computers were large and expensive – Recommend: visit Computer History Museum Major steps: – Big computers became commercial – IBM et al – Digital Equipment et al created ‘mini-computers’ 1971: the first microprocessor appeared – Intel 4004 handled 4 bits at a time, 50kHz bus – A 4004 + a 4001 (ROM & IO) = minimum system 1974: we had the Intel 8008, w/8 bit memory 1975: added the Intel 8080 and the Motorola 6800 7 IEEE Consumer Electronics Society The Problem to Solve IEEE Consumer Electronics Society Atari’s Market Problem They had been leading the arcade game business, starting with Pong They had extended their position with complex arcade games, like Tank They had succeeded competing with Magnavox for home paddle games The clear next move: bring complex games home Atari knew they had competition everywhere 9 IEEE Consumer Electronics Society Implementation Choices Design new custom application specific integrated circuit (ASIC) for each game – Pong for home use is an example Design a programmable system, using a new microprocessor, and an ASIC to drive the screen and the audio In 1974-1975, microprocessors and memory were expensive! 10 IEEE Consumer Electronics Society A bit mapped design Graphic systems at the time would have a bit map or character map. A common graphics terminal would have a 64x16 character array, backed by a ROM for character graphics. – A 1K byte character RAM cost close to $10. A pure bit map would become prohibitively expensive. On a monochrome monitor, 512x256 would be 16K in DRAM. 11 IEEE Consumer Electronics Society The default screen map design Given the processors and memory in mid- 1975, a programmable architecture looked too expensive. A display of 160 x 96 x 2 bits would need a 4K x 8 DRAM array, perhaps $25 dollars. – Not much RAM left for game variables. A processor like the 6800 was quoted around $25 in large quantity. An ASIC would be needed to manage the RAM and generate the display. 12 IEEE Consumer Electronics Society Then MOS Technology changed the analysis IEEE Consumer Electronics Society 6502 introduction A team led by Chuck Peddle left Motorola and teamed up with MOS Technology to build a better microprocessor Big changes: – Replace enhancement mode NMOS by depletion load NMOS – Adjust the register architecture to define better memory addressing – Adjust the architecture for little-endian address arithmetic pipelining 14 IEEE Consumer Electronics Society How this changed the analysis The 6502 was priced at $25 per sample – I bought one in September 1975 at Wescon – So did Wosniak and Ron Milner of Atari MOS Technology quoted $5/each in large quantities Depletion load NMOS made the parts fast enough for 1-2 MHz memory. – Median 6800s could handle 500 kHz memory The architecture changes were decisive 15 IEEE Consumer Electronics Society Address arithmetic Data processing involves manipulating addresses The 8080 had a 16 bit H and L register pair – Computing an address was intensive The 6800 has some address computation – It was limited, a 16 bit X register was offset with an immediate 8 bit value The 6800 had a nice trick: the first 256 ‘zero page’ memory locations could be addressed with shorter instructions 16 IEEE Consumer Electronics Society 6502 memory addressing There were two innovative standard addressing modes for most ALU instructions: Indexed indirect addressing • 8 bit X register added to 8 bit zero page offset. This selects a 16 bit memory address. Indirect indexed addressing • 8 bit zero page offset selects a 16 bit memory address. The 8 bit Y register is added to that 16 bit memory address (next slide). 17 IEEE Consumer Electronics Society Indirect indexed addressing Bus Address bus Data Bus Description clock 1 PC Op code Load instruction using indexed indirect 2 PC+1 ZPA 8 bit value pointing to a base address in zero page 3 ZPA DAL Fetch the Data Address Low byte 4 ZPA + 1 DAH Fetch the Data Address High byte 5 DAH:DAL + Y -data- Add the Y index register to the DAL+DAH to generate the effective address 18 IEEE Consumer Electronics Society Consequence of 6502 speed The 6502 was fast enough to compute and write graphics on a line-by-line basis. IT DID NOT NEED A FRAME BUFFER! – This saved a lot of money. The 6502 could run at 1.2 MHz, for 76 memory cycles per TV horizontal line. – This would be maybe 25 instructions/line. Ron Milner and Steve Mayer of Atari build a prototype (next slide) – Great luck: I was hired to debug it 19 IEEE Consumer Electronics Society Atari VCS proof of concept Computer History Museum 20 IEEE Consumer Electronics Society Generating a Product IEEE Consumer Electronics Society Assembling the team The prototype and I were sent to Los Gatos I was apprenticed to Jay Miner (next slide) Jay worked for Bob Brown, who in turn worked for Al Alcorn – engineering VP We assembled a team to build the system: Wade Tuma, Neils Strohl, etc. We had ASIC chip designers and testers We needed game designers. We hired Larry Wagner. Larry in turn hired Dave Crane, Al Miller, Larry Kaplan and Bob Whitehead. 22 IEEE Consumer Electronics Society Jay Miner 23 IEEE Consumer Electronics Society General plan We looked at the prototype as an existence proof. As Brooks wrote (Mythical Man-Month) we ‘plan to throw the first one away’ We did a new system design, with a CPU, an ASIC, a combo device with memory and IO and an external ROM on a cartridge. – We had decided that having RAM to load a program would cost to much We designed a depletion load NMOS design Then we emulated that design with 74LS TTL 24 IEEE Consumer Electronics Society System fundamentals We kept the essentials: 1. A 6507, a cost reduced 6502 in a 28 pin package, losing some functionality, but saving 50 cents. 2. A 6532, with 128 bytes of RAM, 16 IO ports and a programmable timer 3. The “Television Interface Adaptor”, with video, audio and six extra specialized input ports 4. A 2K or 4K byte ROM on an inserted cartridge. 25 IEEE Consumer Electronics Society Atari VCS system diagram 26 IEEE Consumer Electronics Society Sound circuits Two sound circuits, one per player Three components for each: 1. frequency divider – Start with 2x horizontal rate ~ 31 kHz – Add a 5 bit prescaler 2. noise generator: 4 bit, 5 bit and 9 bit polynomial counters – 4 bit poly counter made tank noises – 9 bit poly counter made jet noises 3. multiplying 4 bit A/D converter 27 IEEE Consumer Electronics Society Parallel IO 5 6532 IO ports for the game console controls: game start, game select, difficulty (left and right) and color/BW. The 6th switch was power. 4 6532 IO ports each for the two controllers 2 TIA paddle port inputs for each of the two controllers; we could play 4 player pong games. 1 TIA trigger port for each of the two controllers; they could also support a light gun 28 IEEE Consumer Electronics Society HW Optimization: collision registers In games, objects ‘collide’ with other objects Detecting collisions in software is difficult. The program needs to construct the outline of each object and compare locations. It was much easier to add a set of 15 hardware collision registers – simple RS FF The programmer reads and clears them. The hardware sets them if any pair of objects and playfield overlap. 29 IEEE Consumer Electronics Society SW Optimization: CHRST device driver Atari designs used low resolution ‘playfield’ and high resolution moving ‘players’ (sprites) Players and ‘missiles’ could have their horizontal positions reset, and be moved a few pixels; they did not have binary position registers and comparators.